A Decision Tree-Based Supplier Procurement Management Method and System

Abstract

The present invention relates to the field of decision tree procurement management, and provides a decision tree-based supplier procurement management method and system, including: S1: Obtain the index information of enterprise suppliers, and integrate and store the index information in a database; S2: Extract a sample data set from the database, and construct a decision tree through the sample data set; S3: Compile expected indicators according to the procurement plan of the supplier, and obtain a decision result through the expected indicators and the decision tree; S4: Determine whether the decision result meets the preset requirements. If it meets, enter step S5; otherwise, output a prompt warning message and end the process; S5: Obtain the sample update amount of the database. If the sample update amount reaches the preset threshold, return to step S2 to reconstruct the decision tree, and prune the current decision tree according to the hit rate set of the historical decision tree; otherwise, return to step S3. The present invention can dynamically refresh the decision tree. As the sample data increases, the decision tree will become more and more accurate, so that the procurement management always maintains a high degree of accuracy.

Description

Technical Field

[0001] The present invention relates to the field of decision tree procurement management, and particularly to a method and system for supplier procurement management based on a decision tree. Background Art

[0002] Scientific decision-making is an important responsibility of modern managers. In the procurement management of enterprise suppliers, the common scenario we encounter is that several procurement plans are formulated, and the internal and external environments of the enterprise are analyzed. Most conditions are known, but there are still certain uncertain factors. Procurement has certain risks, especially in the manufacturing industry, where the quality of supplier procurement directly affects the entire production chain of the enterprise. And the existing procurement management methods cannot update the decision-making system in real time, resulting in the decision-making system being unable to make adjustments synchronously with the changes in samples, and the accuracy of the decision-making system gradually decreasing over time.

[0003] The above content is only used to assist in understanding the technical solution of the present invention, and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides a method for supplier procurement management based on a decision tree, including the steps of:

[0005] S1: Obtain the index information of enterprise suppliers, and integrate and store the index information in a database;

[0006] S2: Extract a sample data set from the database, and construct a decision tree through the sample data set;

[0007] S3: Compile an expected index according to the procurement plan of the supplier, and obtain a decision result through the expected index and the decision tree;

[0008] S4: Determine whether the decision result meets a preset requirement. If it meets, go to step S5; otherwise, output a prompt warning message and end the process;

[0009] S5: Obtain the sample update amount of the database. If the sample update amount reaches a preset threshold, return to step S2 to reconstruct the decision tree, and prune the current decision tree according to the hit rate set of the historical decision tree; otherwise, return to step S3.

[0010] Preferably, step S1 is specifically:

[0011] S11: Output the index information through an SRM (Supplier Relationship Management) system, and the SRM system includes: a material data maintenance unit, a supplier information maintenance unit, a supplier access process unit, a supplier price management unit, a supplier problem management unit, and a supplier order delivery management unit;

[0012] S12: Obtain the source list information and supplier evaluation information from the material data maintenance unit, the supplier information maintenance unit, and the supplier access process unit; obtain the supply price library information from the supplier price management unit; obtain the supplier problem rectification history information from the supplier problem management unit; obtain the supplier order delivery situation information from the supplier order delivery management unit;

[0013] S13: The index information includes: the source list information, the supplier evaluation information, the supply price library information, the supplier problem rectification history information, and the supplier order delivery situation information; store the index information in the database.

[0014] Preferably, step S2 is specifically as follows:

[0015] S21: Construct a table from the sample data set according to the types of discrete attributes, and the types of discrete attributes include: quality score, cost score, delivery score, R & D cycle, purchase price, number of problems, problem severity level, number of purchase orders, order delivery on-time rate, supplier level, supplier status, and whether to purchase;

[0016] S22: Denote the sample data set as D, and denote the proportion of samples with the conclusion of "yes" for whether to purchase as p k , where k represents the count, calculate the information entropy Ent(D) of D, and the calculation formula is as follows:

[0017]

[0018] Among them, y represents the number of conclusions for whether to purchase;

[0019] S23: Extract a certain discrete attribute a from the sample data set D. The discrete attribute a has a total of V value segments {a 1 , a 2 , …, a V}, extract all sample data with values in the value segment a v and denote it as D v , where v represents the number of the value segment of the discrete attribute a, and v takes integer values in the range of 1 to V. Calculate the information entropy Ent(D v ) of D v );

[0020] Obtain the information gain by partitioning the sample data set D through the discrete attribute a, and the calculation formula is as follows:

[0021]

[0022] Among them, |D v | represents the number of values in the value segment av The number of samples within, |D| represents the number of samples in the sample data set;

[0023] S24: Calculate the inherent value IV(a) of the discrete attribute a, and the calculation formula is as follows:

[0024]

[0025] Calculate the information gain ratio Gain_tatio(D,a) of the discrete attribute a, and the calculation formula is as follows:

[0026]

[0027] S25: Repeat steps S22 - S24 to obtain the information gain ratios of all discrete attributes in the sample data set, and construct the decision tree through each of the information gain ratios.

[0028] Preferably, in step S5, the current decision tree is pruned according to the set hit rate of the historical decision trees, specifically:

[0029] S51: Obtain the set of hit rates of the current decision tree, and the expression is:

[0030] Set (q,c) <high,sort,current,up,branch>

[0031] Among them, q represents the number of times the current decision tree is constructed, c represents the types of discrete attributes, high represents the height of the current node in the decision tree, sort represents the position of the current node from the left at the layer height, up represents the category of the upper-level node of the current node, current represents the category of the node corresponding to the current discrete attribute, and branch represents the link branch of the current node;

[0032] S52: Obtain the set of hit rates of the historical decision trees, including:

[0033] Set (1,c) <high,sort,current,up,branch> to Set (q-1,c) <high,sort,current,up,branch>;

[0034] S53: Calculate the hit rate M of the node corresponding to the discrete attribute c in the current decision tree c , and the calculation formula is:

[0035]

[0036] Among them, sum(c) represents the total number of set categories of the discrete attribute c; R cRepresents the number of hits for discrete attribute c;

[0037] S54: Repeat steps S51 - S53 to obtain the hit rates of the nodes corresponding to all discrete attributes in the current decision tree, and prune the nodes with a hit rate lower than 60%.

[0038] A decision tree - based supplier procurement management system, comprising:

[0039] An index information acquisition module, configured to acquire the index information of enterprise suppliers and integrate and store the index information in a database;

[0040] A decision tree construction module, configured to extract a sample data set from the database and construct a decision tree through the sample data set;

[0041] A decision result acquisition module, configured to prepare expected indexes according to the procurement plan of the supplier and obtain a decision result through the expected indexes and the decision tree;

[0042] A decision result judgment module, configured to judge whether the decision result meets a preset requirement. If it meets, enter the decision tree reconstruction module; otherwise, output a prompt warning message and end the process;

[0043] A decision tree reconstruction module, configured to obtain the sample update amount of the database. If the sample update amount reaches a preset threshold, return to the decision tree construction module to reconstruct the decision tree, and prune the current decision tree according to the hit rate set of the historical decision tree; otherwise, return to the decision result acquisition module.

[0044] The present invention has the following beneficial effects:

[0045] 1. Through the decision tree, suppliers that meet all preset conditions can be output efficiently and stably, greatly improving the efficiency of procurement management and reducing the risk of management errors;

[0046] 2. Combining with the data in the SRM database, the decision tree can be dynamically refreshed. As the sample data increases, the decision tree will become more and more accurate, so that the procurement management always maintains a high accuracy. Brief Description of the Drawings

[0047] Figure 1 Is the flowchart of the method of the embodiment of the present invention;

[0048] Figure 2 Is the schematic diagram of the decision tree of the present invention;

[0049] The realization, functional features and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. Detailed Embodiment

[0050] It should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0051] Referring to Figure 1 , the present invention provides a decision tree-based supplier procurement management method, including the steps of:

[0052] S1: Obtain the index information of the enterprise's suppliers, and integrate and store the index information in a database;

[0053] S2: Extract a sample data set from the database, and construct a decision tree through the sample data set;

[0054] S3: Compile expected indexes according to the procurement plans of the suppliers, and obtain decision results through the expected indexes and the decision tree;

[0055] S4: Determine whether the decision result meets the preset requirements. If it meets, go to step S5; otherwise, output a prompt warning message and end the process;

[0056] S5: Obtain the sample update amount of the database. If the sample update amount reaches the preset threshold, return to step S2 to reconstruct the decision tree, and prune the current decision tree according to the hit rate set of the historical decision tree; otherwise, return to step S3.

[0057] In this embodiment, step S1 is specifically as follows:

[0058] S11: Output the index information through the SRM supplier relationship management system, and the SRM supplier relationship management system includes: a material data maintenance unit, a supplier information maintenance unit, a supplier access process unit, a supplier price management unit, a supplier problem management unit, and a supplier order delivery management unit;

[0059] S12: Obtain the source list information and supplier evaluation information from the material data maintenance unit, the supplier information maintenance unit, and the supplier access process unit; obtain the supply price library information from the supplier price management unit; obtain the supplier problem rectification resume information from the supplier problem management unit; obtain the supplier order delivery situation information from the supplier order delivery management unit;

[0060] S13: The index information includes: the source list information, the supplier evaluation information, the supply price library information, the supplier problem rectification resume information, and the supplier order delivery situation information; store the index information in the database.

[0061] In this embodiment, step S2 is specifically as follows:

[0062] S21: Construct a table from the sample data set according to the types of discrete attributes, where the types of discrete attributes include: quality score, cost score, delivery score, R & D cycle, purchase price, number of problems, problem severity level, number of purchase orders, order delivery timeliness rate, supplier level, supplier status, and whether to purchase;

[0063] Specifically, the constructed table is shown in Table 1

[0064] Table 1 Statistical Table of Discrete Attributes

[0065]

[0066] Among them, there are a total of 5 samples;

[0067] S22: The core of the traditional ID3 algorithm is to use information gain as the attribute selection criterion when selecting attributes at each level of the decision tree, so that when testing at each non-leaf node, the maximum class information about the tested records can be obtained; since the number of instances of the sample nodes in the ID3 algorithm may affect the stability of the decision tree, information gain is not a very good feature selection metric; therefore, the present invention uses C4.5, that is, the information gain ratio, as the feature selection metric to construct a decision tree, specifically:

[0068] Denote the sample data set as D, and denote the proportion of samples with the conclusion of "yes" for whether to purchase as p k , k represents the count, calculate the information entropy Ent(D) of D, and the calculation formula is as follows:

[0069]

[0070] Among them, y represents the number of conclusions for whether to purchase;

[0071] S23: Extract a certain discrete attribute a from the sample data set D. The discrete attribute a has a total of V value segments {a 1 , a 2 , …, a v}, extract all sample data with values within the value segment a v in the discrete attribute a and denote it as D v , v represents the number of the value segment of the discrete attribute a, and v takes integer values in the range of 1 to V. Calculate the information entropy Ent(D v ) of D v ;

[0072] Obtain the information gain by dividing the sample data set D through the discrete attribute a, and the calculation formula is as follows:

[0073]

[0074] Among them, |Dv |represents the number of samples with values within the value range segment a v , |D| represents the number of samples in the sample data set;

[0075] S24: Calculate the inherent value IV(a) of the discrete attribute a, and the calculation formula is as follows:

[0076]

[0077] Calculate the information gain ratio Gain_tatio(D, a) of the discrete attribute a, and the calculation formula is as follows:

[0078]

[0079] S25: Repeat steps S22 - S24 to obtain the information gain ratios of all discrete attributes in the sample data set, and construct the decision tree through each of the information gain ratios.

[0080] Specifically, construct a decision tree according to the sample data set in Table 1;

[0081] For the discrete attribute (whether to purchase), the conclusions include: yes and no. Therefore, the value of the number of conclusions y for whether to purchase is 2;

[0082] The proportion of the conclusion of whether to purchase being yes is 3 / 5, and the proportion of the conclusion of whether to purchase being no is 2 / 5;

[0083] Calculate the obtained information entropy:

[0084] Take the calculation of the discrete attribute (quality score) as an example;

[0085] Divide the value range segments of the quality score into: 60 - segment, 70 - segment, and 80 - segment;

[0086] The value range of the samples belonging to the 60 - segment is [60, 70), and the segment number v is 1;

[0087] The value range of the samples belonging to the 70 - segment is [70, 80), and the segment number v is 2;

[0088] The value range of the samples belonging to the 80 - segment is [80, 100], and the segment number v is 3;

[0089] The number of samples with a quality score between [60, 70) is 1, the number of samples with whether to purchase being yes is 0, and the number of samples with whether to purchase being no is 1. Therefore, p k = 0, Ent(D 质量60分段 ) = -(0 + 0) = 0

[0090] The number of samples with quality scores in the range of [70, 80) is 2. The number of "yes" for procurement is 1, and the number of "no" for procurement is 1. Therefore, p k = 1 / 2,

[0091] The number of samples with quality scores in the range of [80, 100] is 2. The number of "yes" for procurement is 2, and the number of "no" for procurement is 0. Therefore, p k = 1, Ent(D 质量80分段 ) = -(0 + 0) = 0;

[0092] The total number of value segments V is 3,

[0093] Calculate the information gain:

[0094]

[0095] Calculate the inherent value of the quality score attribute:

[0096]

[0097] Calculate the information gain ratio of the quality score:

[0098]

[0099] Repeat the above steps to calculate the information gain ratio of each discrete attribute in turn:

[0100] Root node (i.e., information entropy Ent(D)): 0.967, quality score: 0.373, cost score: 0.242, delivery score: 0.186, purchase price: 0.352, number of problems: 0.285, etc.;

[0101] Among them, the information gain ratio of the quality score is the largest, so the first-level decision tree is the quality score. By analogy, the second-level, third-level, etc. of the decision tree can be constructed, and finally a complete decision tree similar to the following figure can be constructed as Figure 2 shown.

[0102] In this embodiment, in step S5, the current decision tree is pruned according to the set hit rate of the historical decision tree. Specifically:

[0103] S51: Obtain the hit rate set of the current decision tree. The expression is:

[0104] Set (q,c) <high,sort,current,up,branch>

[0105] Among them, q represents the number of times the current decision tree is constructed, c represents the types of discrete attributes, high represents the height of the current node in the decision tree, sort represents the position of the current node from the left at the current layer height, up represents the category of the upper-level node of the current node, current represents the category of the node corresponding to the current discrete attribute, and branch represents the link branch of the current node;

[0106] S52: Obtain the hit rate set of the historical decision tree, including:

[0107] Set (1,c) <high, sort, current, up, branch> to Set (q-1,c) <high, sort, current, up, branch>;

[0108] S53: Calculate the hit rate M of the node corresponding to the discrete attribute c in the current decision tree c , and the calculation formula is:

[0109]

[0110] Among them, sum(c) represents the total number of set categories of the discrete attribute c; R c represents the number of hits of the discrete attribute c (that is, the number of times the discrete attribute c remains consistent in all decision trees);

[0111] S54: Repeat steps S51 - S53 to obtain the hit rates of the nodes corresponding to all discrete attributes in the current decision tree, and prune the nodes with a hit rate lower than 60%.

[0112] Specifically, for example, the historical decision tree has been constructed 4 times in total;

[0113] The calculated hit rate of the quality score in the current decision tree is: 1 * (5 / 5) = 100%; the hit rate of the purchase price is: (1 / 3) * (3 / 5) = 20%; then the purchase price is pruned.

[0114] The present invention provides a decision tree-based supplier procurement management system, including:

[0115] An index information acquisition module, configured to acquire index information of enterprise suppliers and integrate and store the index information in a database;

[0116] A decision tree construction module, configured to extract a sample data set from the database and construct a decision tree through the sample data set;

[0117] A decision result acquisition module, configured to compile expected indexes according to the procurement plan of the supplier and obtain a decision result through the expected indexes and the decision tree;

[0118] A decision result judgment module, configured to judge whether the decision result meets a preset requirement. If it meets the requirement, it enters the decision tree reconstruction module; otherwise, it outputs a prompt warning message and ends the process.

[0119] A decision tree reconstruction module, configured to obtain the sample update amount of the database. If the sample update amount reaches a preset threshold, it returns to the decision tree construction module to reconstruct the decision tree, and prunes the current decision tree according to the hit rate set of the historical decision tree; otherwise, it returns to the decision result acquisition module.

[0120] It should be noted that in this article, the terms "including", "comprising" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or system. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of another identical element in the process, method, article or system including that element.

[0121] The serial numbers of the embodiments of the present invention above are only for description and do not represent the superiority or inferiority of the embodiments. Among the unit claims listing several devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order and these words can be interpreted as identifiers.

[0122] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A decision tree-based supplier procurement management method, characterized in that, it includes the steps of: S1: Obtain the index information of the enterprise's suppliers, and integrate and store the index information in the database; S2: Extract a sample data set from the database, and construct a decision tree through the sample data set; S3: Compile expected indexes according to the procurement plan of the supplier, and obtain a decision result through the expected indexes and the decision tree; S4: Determine whether the decision result meets the preset requirements. If it meets, enter step S5; otherwise, output a prompt warning message and end the process; S5: Obtain the sample update amount of the database. If the sample update amount reaches the preset threshold, return to step S2 to reconstruct the decision tree, and prune the current decision tree according to the hit rate set of the historical decision tree; otherwise, return to step S3; In step S5, pruning the current decision tree according to the set hit rate of the historical decision tree is specifically: S51: Obtain the hit rate set of the current decision tree, and the expression is: Set (q,c) <high,sort,current,up,branch> where q represents the construction times of the current decision tree, c represents the types of discrete attributes, high represents the height of the current node in the decision tree, sort represents the position of the current node from the left at the layer height, up represents the category of the upper node of the current node, current represents the category of the node corresponding to the current discrete attribute, and branch represents the link branch of the current node; S52: Obtain the hit rate set of the historical decision tree, including: Set (1,c) <high, sort, current, up, branch> to Set (q-1,c) <high, sort, current, up, branch>; S53: Calculate the hit rate M of the node corresponding to the discrete attribute c in the current decision tree c , and the calculation formula is: Among them, sum(c) represents the total number of set categories of the discrete attribute c; R c represents the number of hits of the discrete attribute c; S54: Repeat steps S51 - S53 to obtain the hit rates of the nodes corresponding to all discrete attributes in the current decision tree, and prune the nodes with a hit rate lower than 60%.

2. The decision tree-based supplier procurement management method according to claim 1, characterized in that, step S1 is specifically: S11: Output the index information through the SRM supplier relationship management system, and the SRM supplier relationship management system includes: a material data maintenance unit, a supplier information maintenance unit, a supplier access process unit, a supplier price management unit, a supplier problem management unit, and a supplier order delivery management unit; S12: Obtain the source list information and supplier evaluation information from the material data maintenance unit, the supplier information maintenance unit, and the supplier access process unit; obtain the supply price library information from the supplier price management unit; obtain the supplier problem rectification resume information from the supplier problem management unit; obtain the supplier order delivery situation information from the supplier order delivery management unit; S13: The index information includes: the source list information, the supplier evaluation information, the supply price library information, the supplier problem rectification resume information, and the supplier order delivery situation information; store the index information in the database.

3. The decision tree-based supplier procurement management method according to claim 1, characterized in that, step S2 is specifically: S21: Construct a table from the sample data set according to the types of discrete attributes, where the types of discrete attributes include: quality score, cost score, delivery score, R & D cycle, purchase price, number of problems, problem severity level, number of purchase orders, order delivery timeliness rate, supplier level, supplier status, and whether to purchase; S22: Denote the sample data set as D, and denote the proportion of samples with the conclusion of "yes" for procurement as p k , where k represents the count. Calculate the information entropy Ent(D) of D, and the calculation formula is as follows: where y represents the number of conclusions on whether to purchase; S23: Extract a certain discrete attribute a from the sample data set D. The discrete attribute a has V value segments {a 1 , a 2 , …, a V}. Extract all the sample data with values of the discrete attribute a within the value segment a v and denote it as D v . Let v represent the number of the value segment of the discrete attribute a, and v takes integer values in the range from 1 to V. Calculate the information entropy Ent(D v ) of D v ; The information gain is obtained by partitioning the sample data set D through the discrete attribute a, and the calculation formula is as follows: where, |D v | represents the number of samples with values within the value range segment a v ; |D| represents the number of samples in the sample data set; S24: Calculate the attribute intrinsic value IV(a) of the discrete attribute a, and the calculation formula is as follows: Calculate the information gain ratio Gain_tatio(D,a) of the discrete attribute a, and the calculation formula is as follows: S25: Repeat steps S22 - S24 to obtain the information gain ratios of all discrete attributes in the sample data set, and construct the decision tree through each of the information gain ratios.

4. A decision tree-based supplier procurement management system for implementing the decision tree-based supplier procurement management method according to any one of claims 1 - 3, characterized in that, it includes: An index information acquisition module for acquiring the index information of the enterprise's suppliers and integrating and storing the index information in a database; A decision tree construction module for extracting a sample data set from the database and constructing a decision tree through the sample data set; A decision result acquisition module for formulating expected indexes according to the procurement plan of the supplier and obtaining a decision result through the expected indexes and the decision tree; A decision result judgment module for judging whether the decision result meets the preset requirements. If it meets, it enters the decision tree reconstruction module; otherwise, it outputs a prompt warning message and ends the process; A decision tree reconstruction module for obtaining the sample update amount of the database. If the sample update amount reaches the preset threshold, it returns to the decision tree construction module to reconstruct the decision tree, and prunes the current decision tree according to the hit rate set of the historical decision tree; Otherwise, it returns to the decision result acquisition module.

Images